Analyzing E-Commerce Data Using Python¶

This data is typical for what you'd see for an ecommerce website. It contains session data with traffic source, location and transcation info. Other data has been anonymized. We will be using Python to analyze this dataset, getting more insights into what is driving sales on this particular day.

E-Commerce

Step 1: Import Packages 📦¶

I'll be using pandas and plotly for this project.

In [1]:
# Import Packages
import pandas as pd
import plotly.express as px

Step 2: Import the Data ⬇️¶

We are going to import the CSV file using pd.read_csv("file.csv")

In [2]:
ecomm_data = pd.read_csv("sample_data - ga_sessions.csv")

ecomm_data.head()
Out[2]:
visitId visitStartTime visitNumber channelGrouping browser operatingSystem deviceCategory continent subContinent country transactions
0 1501591568 1501591568 1 Organic Search Chrome Windows desktop Europe Southern Europe Greece NaN
1 1501589647 1501589647 2 Referral Chrome Windows desktop Asia Southern Asia India NaN
2 1501616621 1501616621 1 Referral Chrome Windows desktop Europe Northern Europe United Kingdom NaN
3 1501601200 1501601200 1 Referral Firefox Windows desktop Americas Northern America United States NaN
4 1501615525 1501615525 1 Referral Chrome Windows desktop Americas Northern America United States NaN

Step 3: Clean the Data 🧹¶

I will be inspecting the type of columns, and make some neccessary changes and adjustments if it is required.

In [3]:
ecomm_data.dtypes
Out[3]:
visitId              int64
visitStartTime       int64
visitNumber          int64
channelGrouping     object
browser             object
operatingSystem     object
deviceCategory      object
continent           object
subContinent        object
country             object
transactions       float64
dtype: object

Changing Datatypes¶

1) I noticed that the 'visitStartTime' column is an int datatype. Let me change it to the datetime datatype.

  • Keep in mind that the unit you specify should match the unit of the timestamps in your data to ensure accurate conversion to datetime objects. (it is up to seconds for this column, hence units = "s")

2) I also noticed that the 'transactions' column is a float datatype. Let me change it to the int datatype.

  • Do check that the column does not have any empty/NaN values, else the datatype conversion will not work.
In [4]:
# It is a good habit to create a copy of the original df and save it into a new df name
ecomm_data_clean = ecomm_data.copy()

# This line of code is converting the values in the 'visitStartTime' column from seconds 
ecomm_data_clean["visitStartTime"] = pd.to_datetime(ecomm_data_clean['visitStartTime'], unit="s")

# Any missing values will be replaced with 0.
ecomm_data_clean["transactions"] = ecomm_data_clean["transactions"].fillna(0)

# The "transactions" column in the ecomm_data_clean DataFrame will contain integer values, 
ecomm_data_clean["transactions"] = ecomm_data_clean["transactions"].astype("int")

ecomm_data_clean
Out[4]:
visitId visitStartTime visitNumber channelGrouping browser operatingSystem deviceCategory continent subContinent country transactions
0 1501591568 2017-08-01 12:46:08 1 Organic Search Chrome Windows desktop Europe Southern Europe Greece 0
1 1501589647 2017-08-01 12:14:07 2 Referral Chrome Windows desktop Asia Southern Asia India 0
2 1501616621 2017-08-01 19:43:41 1 Referral Chrome Windows desktop Europe Northern Europe United Kingdom 0
3 1501601200 2017-08-01 15:26:40 1 Referral Firefox Windows desktop Americas Northern America United States 0
4 1501615525 2017-08-01 19:25:25 1 Referral Chrome Windows desktop Americas Northern America United States 0
... ... ... ... ... ... ... ... ... ... ... ...
2551 1501638264 2017-08-02 01:44:24 1 Referral Chrome Macintosh desktop Americas Northern America United States 0
2552 1501605330 2017-08-01 16:35:30 1 Organic Search Chrome Macintosh desktop Americas Northern America United States 0
2553 1501615026 2017-08-01 19:17:06 3 Organic Search Chrome Macintosh desktop Americas Northern America United States 0
2554 1501627131 2017-08-01 22:38:51 16 Organic Search Chrome Windows desktop Americas Northern America United States 1
2555 1501623575 2017-08-01 21:39:35 2 Referral Chrome Macintosh desktop Americas Northern America United States 2

2556 rows × 11 columns

Taking another look at the datatypes:¶

In [5]:
ecomm_data_clean.dtypes
Out[5]:
visitId                     int64
visitStartTime     datetime64[ns]
visitNumber                 int64
channelGrouping            object
browser                    object
operatingSystem            object
deviceCategory             object
continent                  object
subContinent               object
country                    object
transactions                int64
dtype: object

Both datatypes for 'visitStartTime' and 'transactions' have been successfully changed! 😁

Step 4: Explore the Data 🤔¶

I love to explore my data visually, so let us take a look at it together!

In [6]:
# Learnt to write this function from a professor from school and I've been using it ever since!

def plot_sessions_per_group(df, column_name, viz_type = "bar"): # bar chart will show if viz_type is not specified
    # What this portion does is [1] Group by the Selected Column Name [2] .size() is used to count the number of occurences in each group
    # [3] .reset_index(name = "this is the column name of the counts") [4] .sort_values(by = "this is the column name of the counts")
    sessions_per_group = df.groupby(column_name).size().reset_index(name = "sessions").sort_values(by = "sessions")
    if viz_type == "bar":
            fig = px.bar(sessions_per_group,
                      x = column_name,
                      y = "sessions",
                      title = f'Number of Sessions per {column_name.capitalize()}', #.capitalize() is used to cap the first letter of the word
                      text = "sessions"
                      )
            
            fig.update_yaxes(title_text = "Session Count") # Generate y-axis in plotly (follow the exact syntax)
            fig.update_xaxes(title_text = "Browser") # Generate x-axis in plotly (follow the exact syntax)
            fig.show()

    elif viz_type == "pie":
        return px.pie(sessions_per_group,
                      names = column_name,
                      values = "sessions",
                      title = f"Distribution of Sessions per {column_name.capitalize()}")

    
    else:
        raise ValueError("viz_type can only be 'bar or 'pie'. Please try again!")

Task 1: Find out the number of sessions per browser.¶

In [7]:
plot_sessions_per_group(ecomm_data_clean, "browser")
Insights:¶

1) It appears that Google Chrome is the most commonly used browser, followed by Safari and Firefox

Google Chrome is known for its fast performance, user-friendly interface, and seamless integration with other Google services. As a result, it has become the most popular browser globally, capturing a significant market share.

Task 2: Find out the number of sessions per channelGrouping¶

In [8]:
plot_sessions_per_group(ecomm_data_clean, "channelGrouping", "bar")

Insights: 1) Organic search traffic is considered valuable to e-commerce websites because it typically represents users who are actively seeking out products or services and are more likely to convert into customers.

Task 3: Find out the Distribution of sessions per DeviceCategory¶

In [9]:
plot_sessions_per_group(ecomm_data_clean, "deviceCategory", viz_type="pie")

Insights: 1) It seems that desktop devices are the most common entry point for shopping sessions, followed by mobile devices and then tablets.

Task 4: Find out the number of sessions per Continent¶

In [10]:
plot_sessions_per_group(ecomm_data_clean, "continent", viz_type="bar")

Task 5: Show hourly visits per channelGrouping¶

In [11]:
ecomm_data_clean["visitStartTime"] = ecomm_data_clean["visitStartTime"].dt.tz_localize("America/Los_Angeles")
ecomm_data_clean["hour"] = ecomm_data_clean["visitStartTime"].dt.to_period("H")
visits_per_month = ecomm_data_clean.groupby(["hour"]).size().reset_index(name = "visits")
visits_per_month["hour"] = visits_per_month["hour"].astype(str)

fig = px.line(visits_per_month, x= "hour", y = "visits", title = "Visits per hour throughout the day", markers = True)
fig.update_xaxes(title_text = "Hour of Day")
fig.update_yaxes(title_text = "Count of visits")
fig.show()

/var/folders/ql/7h35fnyd76n8h2vchtfz1fbc0000gn/T/ipykernel_25712/3207906609.py:2: UserWarning:

Converting to PeriodArray/Index representation will drop timezone information.

Task 6: Show Channel Grouping Visits per hour throughout the day¶

In [12]:
ecomm_data_clean["visitStartTime"] = ecomm_data_clean["visitStartTime"].dt.tz_convert("America/Los_Angeles")
ecomm_data_clean["hour"] = ecomm_data_clean["visitStartTime"].dt.to_period("H")
visits_per_month = ecomm_data_clean.groupby(["hour","channelGrouping"]).size().reset_index(name = "visits")
visits_per_month["hour"] = visits_per_month["hour"].astype(str)

fig = px.line(visits_per_month, x= "hour", y = "visits", title = "Channel Grouping Visits per hour throughout the day", color = "channelGrouping", markers = True)
fig.update_xaxes(title_text = "Hour of Day")
fig.update_yaxes(title_text = "Count of visits")
fig.show()

/var/folders/ql/7h35fnyd76n8h2vchtfz1fbc0000gn/T/ipykernel_25712/3845196104.py:2: UserWarning:

Converting to PeriodArray/Index representation will drop timezone information.

Task 7: Which Channel leads to the most transactions per session?¶

In [13]:
channel_grouping_stats = (
    ecomm_data_clean
    .groupby("channelGrouping")
    .agg({"transactions" : "sum", "visitId" : "size"})
    .reset_index()
    )

# Calculating the transaction per session
channel_grouping_stats["transaction_per_session"] = channel_grouping_stats["transactions"] / channel_grouping_stats["visitId"]
channel_grouping_stats.sort_values(by = "transaction_per_session", inplace = True, ascending = False)

px.bar(channel_grouping_stats,
        x = "channelGrouping",
        y = "transaction_per_session",
        title = "Transaction per Session")

Thank you for taking time to view my project! 😊¶